Common Gateway Interface (CGI)

Definition

Programs invoked by the web server
Often act as the interface between the web and legacy apps or databases
- The browser is becoming the universal user interface
The web server is becoming the universal backend; study its architecture.
The most popular programming language for CGI is Perl. Others:
- ASP
- VB
- PHP
- Java
- Shell
These programs run as child processes of the web server (Unix) or as threads (NT).
- The server sets some web-related environment variables
- CGI programs run with the user ID of the web server process (often "nobody" or "apache").
- They are granted and denied file access accordingly.

Web Servers Handle Requests

When a hyperlink is clicked, the browser sends a GET request to a web server. The web server must then decide whether to:
- Copy the hyperlinked file back to the browser, or...
- Run the file (as a program) and send its output to the browser.
How does the server know which to do? The rules in srm.conf tell it.

Lab

Create a file named hello.cgi in your document root directory, containing:

print "Content-Type: text/plain\n\nHello world! (love, Perl)\n";

Note: On Unix, your Perl script needs a "shebang".
Open this file in your browser with a URL starting with http://.
How does your server handle your browser's request? Does it copy or run your Perl script?

Running CGI Scripts

In httpd.conf (on NT) or srm.conf (on Unix), you define the condition(s) under which your server runs files instead of copying them.

Common CGI conditions:
1. If the file is in a particular directory tree (commonly cgi-bin) treat it as a program and run it, or...
2. If the filename has a particular extension (commonly .cgi or .pl) treat it as a program and run it.

To use CGI condition 1, edit httpd.conf (or srm.conf):

# ScriptAlias: This controls which directories contain server scripts.
# Format: ScriptAlias fakename realname

ScriptAlias /cgi-bin/ /usr/local/web/cgi-bin/

Multiple ScriptAlias directives enable users to have personal CGI directories.

To use CGI condition 2, edit srm.conf (on Unix) or httpd.conf (on NT):

# Use CGI outside ScriptAliased directories
# Format: AddHandler action-name ext1

# Define .cgi filename extension for scripts:

AddHandler cgi-script .cgi

You must have a Perl interpreter if your CGi scripts are in Perl.
You must have a shell if your CGI scripts are shell scripts.
You must have a Java virtual machine and class libraries if your CGI scripts are Java.
In general, you must have whatever runtime environment your backend programs require.

Lab
Implement condition 2 and test it on `hello.cgi`. Now, does the server copy or run your Perl script?